1. Field of the Invention
This invention relates to a method and an apparatus for diffusing zinc (Zn) into groups III-V compound semiconductors. The groups III-V compound semiconductor means a semiconductor of a pair of a group III element gallium(Ga), indium(In) or aluminum(Al), and a group V element arsenic(As), phosphorus(P) or antimony(Sb). Bulk single crystal wafers are available for GaAs, InP and GaP. GaAs wafers, InP wafers and GaP wafers are useful as substrates of laser diodes (LDs), light emitting diodes (LEDs), photodiodes (PDs) or other semiconductor devices. Though this invention can be applied to any III-V compound semiconductor wafers, explanation will be done by only citing GaAs and InP.
The III-V compound semiconductor wafers are inherently n-type in many cases. Fabrication of a pn-junction requires epitaxial growth of p-type thin films on the n-type wafer, ion implantation of a p-type impurity, or thermal diffusion of a p-type impurity into the n-type wafer. The epitaxial growth of the p-type films is improper for making localized p-regions through a mask. The ion implantation is not the most suitable, since it requires a large apparatus, a plenty of steps and annealing of the ion implanted wafer, which raise the cost of producing the pn-junction. The thermal diffusion is the most suitable way for making pn-junctions in an n-type wafer. Zinc (Zn) acts as a p-type impurity in GaAs or InP crystals. Magnesium (Mg) and cadmium (Cd) are also p-type impurities in GaAs or InP, but Zn is the most feasible p-impurity for InP or GaAs. Zn-diffusion is one of the most important techniques of fabricating LEDs, LDs, PDs and other semiconductor devices having the group III-V semiconductor substrates. The purpose of the Zn-diffusion is to make localized p-regions on a crystal by diffusion.
Here, the crystal includes substrate crystals and film crystals grown on substrate crystals. A purpose of the present invention is to provide a new Zn-diffusion method and apparatus applicable to a wide wafer. Another purpose of the present invention is to provide a Zn-diffusion method and apparatus of high controllability. A further purpose of the present invention is to provide a Zn-diffusion method and apparatus immune from the use of poisonous materials.
This invention is a version of vapor phase diffusion methods but is different from conventional vapor phase diffusion methods. This invention is rather akin to liquid phase epitaxy (LPE). This invention rather diverts the manner and the device from the liquid phase epitaxy to the Zn-diffusion. Though this invention resembles the liquid phase epitaxy, this invention is essentially a vapor phase diffusion of Zn. Instead of material liquid, a vapor of Zn is filled in a sliding jig. This invention is not epitaxy but diffusion. This invention must be clearly discriminated from the liquid phase epitaxy.
This application claims the priority of Japanese Patent Application No.10-213954(10-213954) filed on Jul. 29, 1998 which is incorporated herein by reference.
2. Description of Prior Art
Impurity diffusion is classified into two categories of vapor phase diffusion and solid phase diffusion by the distinction whether the impurity is supplied from solid phase or vapor phase. There is no concept of "liquid phase diffusion". The solid phase diffusion is a new technology recently proposed by the present Inventors for the first time (Japanese Patent Application No.5-177233, Japanese Patent Laying Open No.7-14791). The solid phase diffusion method has steps of growing a Zn-containing InGaAsP film epitaxially on an n-InP crystal substrate and diffusing Zn from the InGaAsP film into the InP substrate by heat. Since the object InP is protected by the InGaAsP film, P atoms do not escape from the bottom InP substrate in spite of heating in the solid phase diffusion. However, an excess number of steps have been prohibiting the practical use of the solid phase diffusion.
The Zn-diffusion is still actually done exclusively by vapor phase diffusion. The vapor phase diffusion is further classified into two methods. One of the vapor phase diffusion methods is a closed tube method. The other is an open tube method. Both two methods are well known. But only the closed tube method is put into practice on a large scale in the semiconductor industry at present. The open tube method is poorly employed on a small scale in some laboratories, because the open tube method is still suffering from unsolved difficulties. Two methods are explained in detail for clarifying the present state of the art of impurity diffusion.
[A. Closed tube method]
FIG. 14 shows a closed tube method for diffusing Zn into a group III-V semiconductor wafer. A long quartz tube 61 having an open end and a closed end is prepared. An InP wafer (or GaAs wafer) 62 is put on an inner spot near an end 60 of the quartz tube 61. A diffusion source 66 is put on an inner point near the other end 65 of the quartz tube 61. The quartz tube 61 is vacuumed and the open end is sealed by an oxygen-hydrogen flame burner. Sometimes the quartz tube 61 necks in a part 63 containing the solid diffusion source 66. The Zn diffusion source 66 is either a sublimable compound of Zn and As or a sublimable compound of Zn and P. For example, zinc phosphide (ZnP.sub.2), zinc arsenide (ZnAs.sub.2) or so is selected as a material of the Zn-diffusion source, because they satisfy the conditions of inclusion of Zn, sublimability from solid phase to vapor phase and immunity from foreign materials except Zn and the substrate material. This method is called a closed tube method, because the quartz tube is fully closed.
The sealed quartz tube 61 is put into a horizontal furnace having heaters 67 and 68. The furnace heaters 67 and 68 heat the whole of the quartz tube 61 and maintain the Zn-diffusion source 66 at a higher temperature than the wafer 62. The Zn-diffusion source 66 sublimes into vapor at the higher temperature. The vapor flies in the quartz tube to the wafer 62 of GaAs or InP and adheres to the wafer at the lower temperature. The Zn atoms diffuse deeply in the wafer by heat. The diffusion depth in the wafer can be controlled by the temperature and the time. After the determined time has passed, the temperature of the furnace is reduced. When the furnace is cooled to a pertinent temperature, the quartz tube 61 is pulled out of the furnace. The object GaAs wafer (or InP wafer) is taken out by breaking the quartz tube 61. The wafer is provided with pn-junctions by the Zn-diffusion.
FIG. 15 shows an improvement of the closed tube method. A long quartz tube 70 is prepared. A diffusion source 76 is put in at an end 75 of the quartz tube 70. A GaAs wafer (or an InP wafer) 74 is placed in a half-closed short tube 73. The half-closed tube 73 is put in at a middle of the quartz tube 70. A vacuum is formed in the quartz tube 70 and the tube end is sealed by the oxygen-hydrogen flame burner. The closed tube is inserted into a furnace having heaters 79, 80 and 82. The diffusion source 76 is heated to the highest temperature by the heater 79 for subliming the source material. The middle part of the tube is heated at the lowest temperature by the heater 80 for converting the diffusion material vapor into powder and once depositing the powder 78 on the inner wall. Then, the powder 78 is heated for flying to the GaAs wafer 74 for depositing on the wafer. There are some new proposals of the close tube methods other than the method of FIG. 15.
Why the tube must be sealed in the closed tube method? The sealing is required for the necessity of controlling the vapor pressure of the group V element (As or P). The closed space enables the dissolving speed of the diffusion source to uniquely determine the vapor pressure of the group V element. The dissolving speed is determined only by the temperature T of the diffusion source. Namely, in the closed tube, the vapor pressure is controlled only by the temperature T of the diffusion source. Maintaining the balance between the dissolution and the absorption of the group V element on the wafer surface, the close tube method carries the Zn compound in vapor phase from the diffusion source to the wafer, deposits the Zn atoms on the wafer and diffuses the Zn atoms deeply into the wafer.
The time and the temperature determine the depth and the concentration of diffusion. Only the closed tube method among various diffusion methods is practically used on a large scale in the semiconductor industry. The closed tube method has many advantages. The wafers are immune from contamination, because Zn is diffused in a closed space separated from the external environments. A great amount of gas is unnecessary. The wafers are not oxidized. The diffusion is stable. The reliability of diffusion is high in the case of deep diffusion. The closed tube method is a matured technique having a long, rich history. Since it is already an old, ripened technique, it is difficult to cite an original document which describes the typical closed tube method. Instead, some proposals for improvements will be explained now.
1 Japanese Patent Laying Open No.60-53018,"method of diffusing impurities into a compound semiconductor" suggested a new way of vapor phase diffusion of zinc (Zn) into InP. Pointing out a problem of prior diffusion of an excess diffusion speed caused by sealing only an InP wafer and a diffusion source of ZnP.sub.2 or Zn.sub.3 P.sub.2, 1 proposed an addition of a solid phosphorus (P) in the close tube for decreasing the diffusion speed. When the closed quartz tube is heated, the P-vapor pressure is raised by the sublimation of the newly added P solid in the closed tube. The Zn-vapor pressure is suppressed by the P-vapor pressure, since the total pressure is restricted by the temperature. The addition of the P-vapor pressure reduces the diffusion speed through the decrement of the Zn-vapor pressure. The solid P plays the role of retarding the diffusion of Zn.
Why must the closed tube method cut down the diffusion speed? Would the high speed diffusion bring about high throughput? It is, however, wrong. Large heat capacity accompanies the quartz closed tube owing a large length and a big thickness. It takes about 15 minutes to heat the quartz tube up to a temperature between 500.degree. C. and 600.degree. C. in the furnace. But the time of diffusion for making a 2 .mu.m deep p-region is only 10 minutes due to the rapid diffusion. It takes several tens of minutes to cool the furnace for decreasing the temperature to room temperature. Heating and cooling of the whole of the quartz tube require a long time due to the large length and the big thickness.
The large heat capacity allows the quartz tube to change the temperature moderately and continually but forbids the tube from varying temperature rapidly. The sublimation of the diffusion source and the Zn-diffusion start even at the step of rising temperature due to the slow change of temperature. The diffusion still continues even at the step of cooling. The diffusion also occurs at extra steps other than the diffusion step. Since the closed tube method controls the diffusion only by temperature, it is impossible to control the start and the end of the diffusion exactly. Since heating and cooling require a longer time than diffusion, the depth of diffusion cannot be correctly determined. There is another problem of the contamination of the wafer by the Zn, because condensed Zn comes to adhering to the wafer surface at the step of cooling. For overcoming these drawbacks, 1 tried to suppress the extra diffusion accompanying the heating step and the cooling step by supplying the P solid in the quartz tube, raising the P-vapor pressure and decreasing the Zn-vapor pressure.
[B. Open tube method]
A quartz tube having openings at both ends is prepared. The open tube method diffuses Zn into an InP wafer or a GaAs wafer by supplying the InP wafer (or GaAs wafer) into the quartz tube, heating the tube to a pertinent temperature, supplying a Zn-containing metallorganic gas and a As- or P-containing gas, for example, arsine (AsH.sub.3) or phosphine (PH.sub.3) into the open quartz tube. The Zn-containing metallorganic gas is prepared from a metallorganic compound having Zn which is liquid at room temperature, for example, dimethyl zinc (Zn(CH.sub.3).sub.2). The Zn-containing metallorganic gas is made by bubbling the metallorganic compound with hydrogen gas. The Zn-containing gas is introduced into the quartz tube from an opening end and becomes in contact with the heated GaAs (or InP) wafer. The metallorganic gas (e.g.,dimethyl zinc) is dissociated by heat into zinc atoms and hydrocarbons. Zn atoms are adsorbed on the surface of the wafer. Zn atoms cover the surface of the wafer. High temperature gives the wafer a high diffusion coefficient. Zn atoms diffuse from the surface to the inner part along with the inclination of the Zn-concentration.
If the wafer were to be bluntly heated in vacuum, the group V atoms would escape from the surface of the III-V wafer owing to the high dissociation pressure at a high temperature. The open tube method introduces PH.sub.3 gas or AsH.sub.3 gas for heightening the vapor pressure of the group V element in order to forbid the group V atoms from dissociating out of the surface. The high pressure of the group V gas balances the dissociation with the adsorption of the group V atoms on the surface of the wafer. The balance of the open tube method is a dynamic balance in which the flowing gas (PH.sub.3 or AsH.sub.3) protects the wafer from dissociation in stead of perfect equilibrium by the static gas. The open tube method is inferior to the closed tube method in the vapor pressure balance. Since the tube is not sealed, the open tube method, however, can treat far larger wafers than the closed tube method. Possibility of processing a large sized wafer is the most conspicuous feature of the open tube method. Another advantage is the sparing of quartz tubes. Someone considers that the open tube method may excel in controllability, because the gas flows are ruled by valve operations. The open tube method, however, has not been practiced on a large scale in factories of the semiconductor industry yet, but has been adopted only for the purpose of experiments in some universities. For example,
2 T. Tsuchiya, T. Taniwatari, T. Haga, T. Kawano, "High-quality Zn-diffused InP-related materials fabricated by the open-tube technique", 7th International Conference of Indium Phosphide and Related Materials p664 (1995, Sapporo) reported a Zn-diffusion by supplying a mixture gas of hydrogen (H.sub.2), dimethyl-Zn, phosphine (PH.sub.3) to an InGaAsP/InP epitaxial wafer in an MOCVD apparatus. Instead of preparing an inherent open tube diffusion apparatus, the MOCVD apparatus was diverted to an open tube method for diffusion. Since the open tube method requires only a heater and an enclosed space which allow gases to flow, the MOCVD apparatus can be a substitute for the quartz tube in the open tube method. Temporary diversion of the MOCVD apparatus on a small scale can be allowed. However, the MOCVD is an apparatus not for diffusion but for epitaxy. Such a high cost diversion would be forbidden on a large, industrial scale.
3 Japanese Patent Laying Open No.62-143421 "method and apparatus for diffusing an impurity" proposed an improvement of the open tube method. It denied the closed tube method for the reason that the diffusion starts midway of the step of rising temperature. FIG. 16 shows the proposed improvement having a horizontal quartz tube 83 with inlets 85 and 86, and an outlet 87. An InP wafer 84 is put at a spot near the outlet 87 within the quartz tube 83. A Zn-source 88 (Zn.sub.2 P) is laid at another spot near the inlet 86 in the quartz tube 83. An inactive gas is supplied into the tube 83 via the middle inlet 85. The flow of the inactive gas can separate the InP wafer 84 from the Zn-source 88. During the steps of rising temperature (heating step) and decreasing temperature (cooling step), the InP wafer 84 is effectively separated from the Zn-source 88 by blowing the inactive gas into the tube 83 from the middle inlet 85. During the step of diffusion, the flow of the inactive gas is stopped and hydrogen gas is supplied into the tube 83 from the end inlet 86. The hydrogen gas carries the vapor including Zn from the Zn-source 88 to the InP wafer 84. The Zn atoms are adsorbed on the surface of the wafer 84. The high temperature forces the Zn atoms to diffuse into the InP crystal. Operation bars penetrate into the tube through the side valves 89 and 90 for conveying the wafer 84 and the diffusion source 88. The swift change of the gases enables the open tube apparatus to forbid the diffusion from occurring during the cooling step and the heating step. The advocates assert that the open tube method can control exactly the depth of diffusion through the timely control of the gas flow.
The closed tube method is endowed with strong points of controllability of the group V gas pressure, saving of material gases, immunity from contamination and practical achievements. Despite many proposals, only the closed tube method is a practical Zn-diffusion method which has been widely carried out on a large scale in the semiconductor industry. The closed tube method, however, is suffering from a drawback of the difficulty of treating large-sized wafers. Since the closed tube method inserts an object wafer into a quartz tube (ampoule), the quartz tube having an inner diameter larger than the outer diameter of the object wafer should be employed. Not automated manipulators but skilled workers still do all the diffusion steps of inserting a wafer, putting an impurity source in a transparent quartz tube, making the tube vacuous and sealing an open end of the quartz tube by a oxygen-hydrogen flame burner. The formidable difficulty forces the experienced technician to handle the sealing step, excluding the possibility of the automatic sealing by a machine. The high melting point of quartz compels the technician to use the oxygen-hydrogen burner. The sealing operation includes the steps of evacuating the tube by a vacuum pump, softening a part of the quartz tube by the burner flame, narrowing the softened part, shutting the tube at the narrowing part, separating the other part of the quartz tube which is still evacuated by the vacuum pump and rounding the separated end of the part containing the wafer and the diffusion source by the burner. All the steps are done by manual operation of the skilled technician.
An increase of the diameter of the quartz tube raises the difficulty of the vacuum sealing of the quartz tube. One-inch diameter InP wafers have been used so far for making LEDs, LDs, PDs or other devices. But two-inch wafers will be employed for making the devices in near future for enhancing the throughput of the wafer process. If a two-inch diameter InP wafer should be inserted into a 3 mm-thick quartz tube, the outer diameter of the quartz tube would be at least 56 mm. It is extremely difficult even for an expert to seal such a wide quartz tube having a diameter of at least 56 mm. The vacuum sealing of the wide quartz tube requires an exquisite skill of an experienced technician.
The closed tube method has another weak point of the necessity of breaking the transparent, expensive quartz tube for taking the treated wafer out. The broken quartz tube cannot be reused. The broken parts of the expensive quartz tube must be thrown into a garbage pit. It is a waste of expensive natural resources. Further, since the quartz tube is broken down, the fragments are spattered. Some of the fragments adhere to the wafer. Further, the spattered fragments sometimes hurt the wafer.
There is a further drawback in the current closed tube method. It is poor controllability, since the diffusion is controlled only by the temperature. The poor controllability submits the unexpected diffusion occurring even during the (heating) step of rising temperature of the quartz tube. In addition, the undesirable diffusion also takes place even during the (cooling) step of decreasing the temperature. It is difficult to repeat the same profile of temperature change of the heating step, the diffusing step and the cooling step many times. Since the temperature profiles fluctuate every cycle of processes. The poor controllability leads to poor reproductivity of the diffusion depth. The diffusion depths disperse at random, in particular, in the case of shallow diffusion. Another difficulty is undesirable deposition of Zn atoms on the wafer during the cooling step. The closed tube method, therefore, is suffering from the problem of the poor controllability and the problem of the technical difficulty in the case of treating large-sized wafers. A desired diffusion method would be excellent in the controllability of diffusion and the applicability to larger wafers.
On the contrary, the open tube method is more suitable for treating large-sized wafers than the closed tube method. A larger wafer may be treated only by replacing the quartz tube by a larger tube. Since the open tube method does not seal the ends of the reaction tube, this method is immune from the technical difficulty of sealing the quartz tube. The open tube method, however, is plagued by other difficulties. The vapor pressure of the group V gas is unstable, because the group V gas and the Zn-containing gas flow in the tube. The instability of the group V gas may invite the dissociation of the V element atoms from the wafer surface. The open tube method has a more serious drawback. A great amount of the V element gas is supplied into the tube for maintaining the V gas pressure. The V element gas, for example, arsine (AsH.sub.3) or phosphine (PH.sub.3), is a strong poison. Protection of the environments would require a large-scaled depollution equipment of the exhaustion gas for the open tube method. The open tube method needs a highly expensive, large apparatus on an industrial scale. Thus, the semiconductor industry has not yet employed the well-known open tube method as Zn-diffusion technology.